Human-caused ignition pathways under climate change scenarios in Eastern Spain

Wildfires pose an increasing threat to society, requiring appropriate approaches to understand the components of risk to design effective mitigation strategies. Under this premise, we present a comprehensive methodology to assess the probability of ignition of human-caused wildfires, one of the key drivers of risk. Our approach combines historical ignition records of fires larger than 5 ha (849 ignitions during 1998–2016) in eastern Spain and geospatial information regarding ignition variables. The method leverages the Random Forest algorithm to train a spatially-explicit model of ignition probability, combining distance to wildland interfaces and roads, population density, fuel types, and daily estimates of dead fuel moisture content (DFMC). The model was applied to outline the spatial pattern of probability under current conditions (2015–2020) and future projections across four Shared Socioeconomic Pathways (SSP1-2.6, 2-4.5, 3-7.0, and 5–8.5). The model achieved satisfactory predictive performance (AUC = 0.76 ± 0.01). We observed a generalized increase in the probability of ignition in all scenarios linked to climate warming decreasing DFMC, except in SSP1-2.6. Furthermore, changes in population density fostered an increase in probability in rural and mountainous areas. Taken together, our findings make an important contribution to fire risk assessment and the development of adaptation strategies under different socioeconomic trends.